Method and apparatus for recognition of character string in scene image

• US 20060039605A1
• Filed: 02/04/2005
• Published: 02/23/2006
• Est. Priority Date: 08/20/2004
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Koga, Masashi
• Granted Patent: US 7,403,656 B2
• Time in Patent Office: Days
• US Class Current: 382/182
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/162 Quantising the image signal

Abstract

A character recognition method that is robust under an unknown illumination condition is provided. An apparatus for realizing such robust character recognition includes plural different binarization, means for synthesizing character sub-image candidates that have been obtained from the binarization units, and means for analyzing character sub-image candidates and for recognizing an image as a character string consisting of character sub-image candidates.

Claims

1. A method for recognition of a character string in a scene image, comprising:

binarization means for obtaining a plurality of mutually different binary images from an input image;

pre-segmentation means for detecting a set of pixels as a character subimage candidate that is assumed as a character from each binary image obtained;

means for detecting a pair of character subimage candidates that are derived from the same binary image or different binary images and are located close to each other from a set of the character subimage candidates obtained by the character subimage candidate detecting means;

means for performing character recognition on the character subimage candidates; and

means for obtaining a combination of the characters that is most plausible as a character string by concatenating character subimage candidates that are adjacent to each other.

2. The method for recognition of a character string in a scene image according to claim 1, further comprising:

means for detecting edges of a plane on which characters are written from an input multi-level or color image; and

means for selecting a character string that is most plausible as a character string based on consistency between obtained edges and layout of the characters.

3. An apparatus for recognition of a character string in a scene image, comprising:

means for taking a picture of a scene and taking it in as an image;

means for storing the image; and

an arithmetic unit for performing processing of the stored image;

wherein,

the arithmetic unit has capabilities of;

obtaining a plurality of mutually different binary images from the input image;

detecting a set of pixels that is assumed as a character character subimage candidate from each binary image obtained;

detecting a pair of character subimage candidates that are derived from the same binary image or different binary images and which are disposed close to each other from a set of character subimage candidates obtained by the character subimage candidate detecting;

performing character recognition on the character subimage candidates; and

obtaining a most plausible combination of the character subimage candidates as a character string by concatenating adjacent character subimage candidates.

4. A program for recognition of a character string in a scene image, comprising:

a plurality of different means for obtaining a plurality of binary images from an input multi-level or color image;

means for detecting a character subimage candidate as a most plausible set of pixels as a character from each binary image;

means for detecting a pair of character subimage candidates that are derived from the same binary image or different binary images and are disposed close to each other from a set of character subimage candidates;

means for performing character recognition on the character subimage candidates; and

means for obtaining a most plausible combination of characters that is plausible as a character string by concatenating character subimage candidates that are adjacent to each other.

5. An apparatus for recognition of a character string in a scene image, comprising:

a plurality of different binarization means for obtaining a plurality of different binary images from an input image;

pre-segmentation means for detecting a set of pixels as a character subimage candidate that is assumed as a character from each binary image obtained;

network synthesizing means for forming a single segmentation hypothesis network from character subimage candidates obtained by the pre-segmentation means; and

character string recognizing means that selects a character subimage candidate that is most plausible as a character from among character subimage candidates of the segmentation hypothesis network, links these character subimage candidates on the hypothesis network, and extracts it as a character string.

6. The apparatus for recognition of a character string in a scene image according to claim 5, further comprising:

text plane detecting means for detecting information of edges of a text plane from the input image; and

candidate character string selecting means for extracting a character string satisfying predetermined conditions using information of the edges of the text plane when a plurality of character string candidates are extracted as the character string.